Tidy tuesday - Unemployment EDA

1 Options & libs

CSS for scrollable output & Header colors

Turning scientific / Exponential numbers off

options(scipen = 999)

library(tidyverse)
library(tidytuesdayR)
library(ggthemes)
library(glue)
library(scales)

Creating & setting custom theme

theme_viny_bright <- function(){
  
  library(ggthemes)
  
  ggthemes::theme_fivethirtyeight() %+replace%
  
  theme(
    axis.title = element_text(size = 9),
    axis.text = element_text(size = 8),
    legend.text = element_text(size = 7),
    panel.background = element_rect(fill = "white"),
    plot.background = element_rect(fill = "white"),
    strip.background = element_blank(),
    legend.background = element_rect(fill = NA),
    legend.key = element_rect(fill = NA),
    plot.title = element_text(hjust = 0.5,
                              size = 16,
                              face = "bold"),
    plot.subtitle = element_text(hjust = 0.5, size = 10, face = "bold"),
    plot.caption = element_text(hjust = 1, size = 8)
      )
  
  }

theme_set(theme_viny_bright())

sources:

Inspired from: https://www.youtube.com/watch?v=gkZ5n8sfXns

2 Loading data

tt <- tt_load("2021-02-23")

    Downloading file 1 of 2: `earn.csv`
    Downloading file 2 of 2: `employed.csv`

tt

3 EDA

employed <- tt$employed
employed

str(employed)

spec_tbl_df [8,184 x 7] (S3: spec_tbl_df/tbl_df/tbl/data.frame)
 $ industry        : chr [1:8184] "Agriculture and related" "Agriculture and related" "Agriculture and related" "Agriculture and related" ...
 $ major_occupation: chr [1:8184] "Management, professional, and related occupations" "Management, professional, and related occupations" "Service occupations" "Service occupations" ...
 $ minor_occupation: chr [1:8184] "Management, business, and financial operations occupations" "Professional and related occupations" "Protective service occupations" "Service occupations, except protective" ...
 $ race_gender     : chr [1:8184] "TOTAL" "TOTAL" "TOTAL" "TOTAL" ...
 $ industry_total  : num [1:8184] 2349000 2349000 2349000 2349000 2349000 ...
 $ employ_n        : num [1:8184] 961000 58000 13000 94000 12000 96000 931000 10000 33000 42000 ...
 $ year            : num [1:8184] 2020 2020 2020 2020 2020 2020 2020 2020 2020 2020 ...
 - attr(*, "spec")=
  .. cols(
  ..   industry = col_character(),
  ..   major_occupation = col_character(),
  ..   minor_occupation = col_character(),
  ..   race_gender = col_character(),
  ..   industry_total = col_double(),
  ..   employ_n = col_double(),
  ..   year = col_double()
  .. )

summary(employed)

   industry         major_occupation   minor_occupation   race_gender       
 Length:8184        Length:8184        Length:8184        Length:8184       
 Class :character   Class :character   Class :character   Class :character  
 Mode  :character   Mode  :character   Mode  :character   Mode  :character  
                                                                            
                                                                            
                                                                            
                                                                            
 industry_total        employ_n             year     
 Min.   :   18000   Min.   :       0   Min.   :2015  
 1st Qu.:  767250   1st Qu.:    9000   1st Qu.:2016  
 Median : 2484000   Median :   65000   Median :2018  
 Mean   : 5077105   Mean   :  461552   Mean   :2018  
 3rd Qu.: 7643000   3rd Qu.:  373000   3rd Qu.:2019  
 Max.   :35894000   Max.   :20263000   Max.   :2020  
 NA's   :660        NA's   :660                      

employed %>% 
  mutate_if(is.character, as.factor) %>% 
  summary()

                          industry   
 Agriculture and related      : 396  
 Construction                 : 396  
 Durable goods                : 396  
 Education and health services: 396  
 Financial activities         : 396  
 (Other)                      :5874  
 NA's                         : 330  
                                                     major_occupation
 Management, professional, and related occupations           :1488   
 Natural resources, construction, and maintenance occupations:2232   
 Production, transportation, and material moving occupations :1488   
 Sales and office occupations                                :1488   
 Service occupations                                         :1488   
                                                                     
                                                                     
                                          minor_occupation
 Construction and extraction occupations          : 744   
 Farming, fishing, and forestry occupations       : 744   
 Installation, maintenance, and repair occupations: 744   
 Office and administrative support occupations    : 744   
 Production occupations                           : 744   
 Professional and related occupations             : 744   
 (Other)                                          :3720   
                    race_gender   industry_total        employ_n       
 Asian                    :1254   Min.   :   18000   Min.   :       0  
 Black or African American:1386   1st Qu.:  767250   1st Qu.:    9000  
 Men                      :1386   Median : 2484000   Median :   65000  
 TOTAL                    :1386   Mean   : 5077105   Mean   :  461552  
 White                    :1386   3rd Qu.: 7643000   3rd Qu.:  373000  
 Women                    :1386   Max.   :35894000   Max.   :20263000  
                                  NA's   :660        NA's   :660       
      year     
 Min.   :2015  
 1st Qu.:2016  
 Median :2018  
 Mean   :2018  
 3rd Qu.:2019  
 Max.   :2020  
               

looks like we have NA’s in data

3.1 Missing Values

sapply(employed, function(x) sum(is.na(x))) %>% 
  as.data.frame()

library(naniar)

employed %>% 
  naniar::gg_miss_upset()

3.2 Cols Freq

table(employed$industry) %>% 
  as.data.frame() %>% 
  arrange(desc(Freq)) %>% 
  
  ggplot(aes(Freq, fct_reorder(Var1, Freq), fill = Var1))  +
  geom_col() +
  theme(legend.position = "none")

top_freq_elements <- function(x){
  
  table(x) %>% 
    as.data.frame() %>% 
    arrange(desc(Freq)) %>% 
    
    ggplot(aes(Freq, fct_reorder(x, Freq), fill = x))  +
    geom_col() +
    theme(legend.position = "none")
}

employed %>% 
  select_if(is.character) %>% 
  map(., .f = top_freq_elements)

$industry

$major_occupation

$minor_occupation

$race_gender

employed %>% 
  count(year)

employed %>% 
  count(industry) %>% 
  ggplot(aes(x = n, y = industry, fill = industry)) +
  geom_col() +
  theme(legend.position = "none")

from: onenote:///\\VINY-PC\Users\viny\Documents\OneNote%20Notebooks\R%20Learning%20&%20Notes\R%20Visualization.one#count()%20%20plot%20frequency%20of%20each%20variable%20in%20a%20function&section-id={C245D183-2D71-46D1-BCBE-2C1A047C220B}&page-id={FEB1F723-D8B8-4DB8-9BA3-ACD8DF96F454}&object-id={A46CDF35-1F4D-45E3-A63B-A89DC402039A}&10

var_freq_plot_fn <- function(df, selected_var){
  df %>% 
    # select_if(is.character) %>%
    count(.data[[selected_var]]) %>%
    # as_tibble() %>%
    
    ggplot(aes(x = n, y = .data[[selected_var]], fill = .data[[selected_var]])) +
    geom_col() +
    theme(legend.position = "none")
}

purrr::map(.x = names(employed %>% select_if(is.character)), 
           .f = var_freq_plot_fn, 
           df = employed)

[[1]]

[[2]]

[[3]]

[[4]]

var_freq_plot_fn <- function(df){
  purrr::map(df %>% 
               select_if(is.character) %>% 
               names, ~
  df %>% 
    count(.data[[.x]]) %>%
    ggplot(aes(x = n, y = .data[[.x]], fill = .data[[.x]])) +
    geom_col() +
    theme(legend.position = "none"))
}

var_freq_plot_fn(df = employed)

[[1]]

[[2]]

[[3]]

[[4]]

3.3 sankey highcharter

from: https://www.youtube.com/watch?v=k-IN6HBhgq4&t=142s

library(highcharter)

highcharter::data_to_sankey(data = (employed %>% 
                                      select(industry, major_occupation))) %>% 
  hchart(., "sankey", name = "Industries to Occupation")

highcharter::data_to_sankey(data = (employed %>% 
                                      select(major_occupation, minor_occupation))) %>% 
  hchart(., "sankey", name = "Major occupation to minor occupation") %>% 
  hc_add_theme(hc_theme_monokai()) %>% 
  hc_title(text = "Industry to Occupation sankey plot")

3.4 Employment Change

employed %>% 
  na.omit() %>% 
  group_by(year) %>% 
  summarise(employment_yrwise = sum(employ_n))

employed %>% 
  na.omit() %>% 
  group_by(year) %>% 
  summarise(employment_yrwise = sum(employ_n)) %>% 
  mutate(employ_change = (employment_yrwise - lag(employment_yrwise, default = 0))/
           lag(employment_yrwise) ) %>% 
  mutate(employ_change = replace(employ_change, is.na(employ_change), 0))

employed %>% 
  na.omit() %>% 
  group_by(year) %>% 
  summarise(employment_yrwise = sum(employ_n)) %>% 
  mutate(employ_change = (employment_yrwise - lag(employment_yrwise, default = 0))/
           lag(employment_yrwise) ) %>% 
  mutate(employ_change = replace(employ_change, is.na(employ_change), 0),
         line_color = ifelse(employ_change >= 0, "blue","red")) %>% 
  
  ggplot(aes(x = year, y = employ_change, 
             label = round(employ_change*100, digits = 2)
             ,col = line_color
             )) +
  geom_line(group=1) +
  geom_point() +
  scale_y_continuous(labels = scales::percent_format(),
                     limits = c(-0.08, 0.02) ) +
  scale_color_identity() +
  geom_text(nudge_y = .005) +
  labs(title = "Yearly % Change in Employment")

from: https://stackoverflow.com/questions/44947806/how-can-i-fill-the-space-between-valuesgeom-line-and-an-intercept-with-ggplot2/44948631#44948631

employment_yr_change <- employed %>% 
  na.omit() %>% 
  group_by(year) %>% 
  summarise(employment_yrwise = sum(employ_n)) %>% 
  mutate(employ_change = (employment_yrwise - lag(employment_yrwise, default = 0))/
           lag(employment_yrwise) ) %>% 
  mutate(employ_change = replace(employ_change, is.na(employ_change), 0))  
  
  
employment_yr_change %>% 
  ggplot(aes(x = year, y = employ_change, 
             label = round(employ_change*100, digits = 2)
             # ,col = line_color
             )) +
  # geom_line(group=1) +
  # geom_point() +
  geom_ribbon(aes(ymin=pmin(employment_yr_change$employ_change,0), ymax=0), fill="red", col="red", alpha=0.5) +
  geom_ribbon(aes(ymin=0, ymax=pmax(employment_yr_change$employ_change,0)), fill="green", col="green", alpha=0.5) +
  
  scale_y_continuous(labels = scales::percent_format(),
                     limits = c(-0.08, 0.02) ) +
  scale_color_manual(values = c("blue","red")) +
  geom_text(nudge_y = .005) +
  labs(title = "Yearly % Change in Employment")

3.4.1 distinct color line

Coloring positive & negative line with separate colors

from: https://stackoverflow.com/questions/66370513/how-to-color-geom-line-geom-point-properly-based-on-condition-if-less-than/66371746#66371746

create function

divide_line <- function(x, y, at = 0) {
  df <- data.frame(x, ymin = at, ymax = y)
  df$sign <- sign(df$ymax - df$ymin)
  df <- df[order(df$x), ]
  df$id <- with(rle(df$sign), rep.int(seq_along(values), lengths))
  
  crossover <- which(c(FALSE, diff(df$id) == 1))
  crossover <- sort(c(crossover, crossover - 1))
  splitter  <- rep(seq_len(length(crossover) / 2), each = 2)
  crossover <- lapply(split(df[crossover, ], splitter), find_isect)
  
  df <- do.call(rbind, c(list(df), crossover))
  df[order(df$x),]
}
find_isect <- function(df) {
  list2env(df, envir = rlang::current_env())
  dx <- x[1] - x[2]
  dy <- ymin[1] - ymin[2]
  t <- (-1 * (ymin[1] - ymax[1]) * dx) / (dx * (ymax[1] - ymax[2]) - dy * dx)
  df$x <- x[1] + t * -dx
  df$ymin <- df$ymax <- ymin[1] + t * -dy
  return(df)
}

create df

df <- divide_line(employment_yr_change$year, employment_yr_change$employ_change, at = 0)

df

create plot

ggplot(df, aes(x, ymax, group = id, colour = as.factor(sign),
               label = paste0(round(ymax*100, digits = 2),"%") )
       )+
  geom_line(size = .9) +
  geom_point() +
  scale_y_continuous(labels = percent_format(), 
                     limits = c(-.08,.02)
                     ) +
  geom_text(nudge_y = .005) +
  labs(title = "Yearly % Change in Employment")

ggplot(df, aes(x, ymax, group = id, colour = as.factor(sign),
               label = paste0(round(ymax*100, digits = 2),"%") )
       )+
  geom_line(size = .9, aes(linetype = as.factor(id))) +
  geom_point() +
  scale_linetype_manual(values=c(2, 1, 3)) +
  scale_y_continuous(labels = percent_format(), 
                     limits = c(-.08,.02)
                     ) +
  geom_text(nudge_y = .005) +
  labs(title = "Yearly % Change in Employment")

3.5 Emp comparison charts

balloon plot from: http://www.sthda.com/english/articles/32-r-graphics-essentials/129-visualizing-multivariate-categorical-data/

https://rpkgs.datanovia.com/ggpubr/reference/ggballoonplot.html

library(ggpubr)
library(viridis)

ggpubr::ggballoonplot(employed, x = "major_occupation", y = "industry", 
                      size = "employ_n", fill = "employ_n") +
  scale_fill_viridis_c(option = "C")

removing irrelevant categories from industries

ggpubr::ggballoonplot(employed %>% 
                      filter(!industry %in% c(NA, "Men","Women","White","Black or African American", "Asian")) %>% 
                        mutate(major_occupation = str_wrap(major_occupation, width = 25),
                               industry = str_wrap(industry, width = 25)), 
                      x = "major_occupation", y = "industry", 
                      size = "employ_n", shape = 16) +
  scale_fill_viridis_c(option = "C") +
  labs(title = "Industry wise Employment Comparison")

ggpubr::ggballoonplot(employed %>% 
                      filter(!industry %in% c(NA, "Men","Women","White","Black or African American", "Asian")) %>%  
                        mutate(industry = str_wrap(industry, width = 25)), 
                      x = "year", y = "industry", 
                      size = "employ_n", fill = "employ_n") +
  scale_fill_viridis_c(option = "C") +
  labs(title = "Industry Employment Comparison year wise")

ggpubr::ggballoonplot(employed %>% 
                      filter(!industry %in% c(NA, "Men","Women","White","Black or African American", "Asian")), 
                      x = "major_occupation", y = "industry", 
                      size = "employ_n", fill = "employ_n",  shape = 21, 
                      facet.by = "year", ggtheme = theme_bw()) +
  
  scale_fill_viridis_c(option = "C") +
  # gradient_fill(c("blue", "white", "red"))
  
  labs(title = "Yearly Industry & occupation wise employment comparison")

employed$minor_occupation %>% unique()

 [1] "Management, business, and financial operations occupations" 
 [2] "Professional and related occupations"                       
 [3] "Protective service occupations"                             
 [4] "Service occupations, except protective"                     
 [5] "Sales and related occupations"                              
 [6] "Office and administrative support occupations"              
 [7] "Farming, fishing, and forestry occupations"                 
 [8] "Construction and extraction occupations"                    
 [9] "Installation, maintenance, and repair occupations"          
[10] "Production occupations"                                     
[11] "Transportation and material moving occupations"             
[12] "Manage-ment, business, and financial operations occupations"

ggpubr::ggballoonplot(employed, x = "major_occupation", y = "minor_occupation", 
                      size = "employ_n", fill = "employ_n",  shape = 21, 
                      facet.by = "year", ggtheme = theme_bw()) +
  
  scale_fill_viridis_c(option = "C") +
  # gradient_fill(c("blue", "white", "red"))
  
  labs(title = "Yearly occupation wise employment comparison")

As we can see in above chart there are two Management, Manage-ment & needs data cleaning and re plotting

employed <- employed %>% 
  mutate(minor_occupation = str_replace(employed$minor_occupation, "Manage-ment", "Management")) 

employed$minor_occupation %>% unique()

 [1] "Management, business, and financial operations occupations"
 [2] "Professional and related occupations"                      
 [3] "Protective service occupations"                            
 [4] "Service occupations, except protective"                    
 [5] "Sales and related occupations"                             
 [6] "Office and administrative support occupations"             
 [7] "Farming, fishing, and forestry occupations"                
 [8] "Construction and extraction occupations"                   
 [9] "Installation, maintenance, and repair occupations"         
[10] "Production occupations"                                    
[11] "Transportation and material moving occupations"            

ggpubr::ggballoonplot(employed, x = "major_occupation", y = "minor_occupation", 
                      size = "employ_n", fill = "employ_n",  shape = 21, 
                      facet.by = "year", ggtheme = theme_bw()) +
  
  scale_fill_viridis_c(option = "C") +
  # gradient_fill(c("blue", "white", "red"))
  
  labs(title = "Yearly occupation wise employment comparison")

3.6 Adding dimension

employed <- employed %>% 
  mutate(dimension = case_when(race_gender == "TOTAL" ~ "Total",
                               race_gender %in% c("Men", "Women") ~ "Gender",
                               TRUE ~ "Race")) 

employed %>% 
  select(dimension) %>% 
  table()

.
Gender   Race  Total 
  2772   4026   1386 

3.7 Industry Emp Comparison

employed_ind_cleaned <- employed %>% 
  na.omit() %>% 
  filter(dimension == "Total") %>% 
  mutate(industry = fct_lump(industry, 11, w = employ_n),
         industry = fct_reorder(industry, employ_n, sum))

3.7.1 Industry Stack plot

employed_ind_cleaned %>% 
  
  ggplot(aes(x = year, y = employ_n, fill = industry)) +
  geom_col() +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme(legend.position = "right", legend.direction = "vertical") +
  labs(title = "Number of employs industry wise across the years") +
  scale_fill_pander()

3.7.2 Industry circular bar

employed_ind_cleaned %>% 
  filter(year == 2020) %>% 
  # summarise(max(employ_n))
  
  ggplot(aes(x = industry, y = employ_n, fill = industry)) +
  geom_bar(stat = "identity") +
  # scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme(legend.position = "none") +
  ylim(0,25000000) +
  coord_polar(start = 0) +
  labs(title = "Number of employs industry wise in 2020")

employed_ind_cleaned %>% 
  filter(year == 2020) %>% 
  
  ggplot(aes(x = fct_reorder(industry, employ_n, sum), y = employ_n, fill = industry)) +
  geom_col() +
  # scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6),
  #                    limits = -50,20000000) +
  theme_minimal() +
  theme(legend.position = "none") +
  ylim(-5000000,25000000) +
  coord_polar(start = 240) +
  labs(title = "Number of employs industry wise in 2020")

3.7.3 Industry facet

employed_ind_cleaned %>% 
  
  ggplot(aes(x = year, y = employ_n, fill = industry)) +
  geom_col() +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme(legend.position = "none",
        strip.text = element_text(size=8, face = "bold")) +
  labs(title = "Number of employs industry wise across the years",
       subtitle = "Keeping scale fixed for industry level comparison") +
  facet_wrap(~ industry) +
  scale_fill_pander()

employed_ind_cleaned %>% 
  
  ggplot(aes(x = year, y = employ_n, fill = industry)) +
  geom_col() +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme(legend.position = "none",
        strip.text = element_text(size=8, face = "bold")) +
  labs(title = "Number of employs industry wise across the years",
       subtitle = "Free scale comparison") +
  facet_wrap(~ industry, scales = "free_y") +
  scale_fill_pander() 

employed_ind_cleaned %>% 
  group_by(industry, year) %>% 
  summarise(employment_yrwise = sum(employ_n)) %>% 
  mutate(employ_change = (employment_yrwise - lag(employment_yrwise, default = 0)),
         industry = str_wrap(industry, width = 20)) %>% 
  
  ggplot(aes(x = year, y = employ_change, col = industry)) +
  geom_line(size = .9) +
  scale_y_continuous(limits = c(-4000000,2000000),
                     labels = unit_format(unit = "M", scale = 1e-6)) +
  facet_wrap(~ industry) +
  
  labs(title = "Yearly Actual Change in employment count Industry wise") +
  theme(legend.position = "none", panel.grid.major = element_blank(),
        strip.text = element_text(size=8, face = "bold")) +
  guides(x = guide_axis(n.dodge = 3))

3.7.4 Industry % change

employed_ind_cleaned %>% 
  group_by(industry, year) %>% 
  summarise(employment_yrwise = sum(employ_n)) %>% 
  mutate(employ_change = (employment_yrwise - lag(employment_yrwise, default = 0))/
           lag(employment_yrwise),
         industry = str_wrap(industry, width = 20) ) %>% 
  mutate(employ_change = replace(employ_change, is.na(employ_change), 0) %>% round(digits = 4)) %>% 
  
  ggplot(aes(x = year, y = employ_change, fill = industry)) +
  geom_col() +
  scale_y_continuous(labels = percent_format()) +
  facet_wrap(~ industry) +
  
  labs(title = "Yearly % Change in employment count Industry wise") +
  theme(legend.position = "none",
        strip.text = element_text(size=8, face = "bold"))

ggsave("emp_change_industry.jpg")

employed_ind_cleaned %>% 
  group_by(industry, year) %>% 
  summarise(employment_yrwise = sum(employ_n)) %>% 
  mutate(employ_change = (employment_yrwise - lag(employment_yrwise, default = 0))/
           lag(employment_yrwise),
         industry = str_wrap(industry, width = 20) ) %>% 
  mutate(employ_change = replace(employ_change, is.na(employ_change), 0) %>% round(digits = 4)) %>% 
  
  ggplot(aes(x = year, y = employ_change, fill = industry)) +
  geom_col() +
  coord_flip() +
  scale_y_continuous(labels = percent_format()) +
  facet_wrap(~ industry) +
  
  labs(title = "Yearly % Change in employment count Industry wise") +
  theme(legend.position = "none",
        strip.text = element_text(size=8, face = "bold"))

ggsave("emp_change_industry_flipped.jpg")

employed_ind_cleaned %>% 
  group_by(industry, year) %>% 
  summarise(employment_yrwise = sum(employ_n)) %>% 
  mutate(employ_change = (employment_yrwise - lag(employment_yrwise, default = 0))/
           lag(employment_yrwise),
         industry = str_wrap(industry, width = 20) ) %>% 
  mutate(employ_change = replace(employ_change, is.na(employ_change), 0) %>% round(digits = 4)) %>% 
  
  ggplot(aes(x = year, y = employ_change, col = industry)) +
  geom_line(size=.9) +
  scale_y_continuous(labels = percent_format()) +
  facet_wrap(~ industry) +
  
  labs(title = "Yearly % Change in employment count Industry wise") +
  theme(legend.position = "none", 
        panel.grid.major = element_blank(),
        strip.text = element_text(size=8, face = "bold"),
        axis.text.x = element_text(angle = 90))

from: https://www.youtube.com/watch?v=_7J6BbDgqrA

but is not working as expected

employed_ind_cleaned %>% 
  group_by(industry, year) %>% 
  summarise(employment_yrwise = sum(employ_n)) %>% 
  mutate(employ_change = (employment_yrwise - lag(employment_yrwise, default = 0))/
           lag(employment_yrwise),
         industry = str_wrap(industry, width = 20) ) %>% 
  mutate(employ_change = replace(employ_change, is.na(employ_change), 0) %>% round(digits = 4)) %>% 
  
  ggplot() +
  geom_col(aes(x = year, y = employ_change * 20000000, fill = employ_change > 0), alpha = 0.4) +
  geom_line(aes(x = year, y = employment_yrwise), group =1) +
  geom_point(aes(x = year, y = employment_yrwise, col = employ_change > 0)) +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  facet_wrap(~ industry) +
  
  labs(title = "Yearly % Change in employment count Industry wise") +
  theme(legend.position = "none", panel.grid.major = element_blank(),
        strip.text = element_text(size=8, face = "bold")) +
  guides(x = guide_axis(n.dodge = 3))

by adding geom_text this worked

employed_ind_cleaned %>% 
  group_by(industry, year) %>% 
  summarise(employment_yrwise = sum(employ_n)) %>% 
  mutate(employ_change = (employment_yrwise - lag(employment_yrwise))/
           lag(employment_yrwise) ) %>% 
  na.omit() %>%
  mutate(industry = str_wrap(industry, width = 20)) %>% 
   
  
  ggplot() +
  geom_col(aes(x = year, y = employ_change * 20000000, fill = employ_change > 0), alpha = 0.4) +
  geom_line(aes(x = year, y = employment_yrwise), group =1) +
  geom_point(aes(x = year, y = employment_yrwise, col = employ_change > 0)) +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  
  facet_wrap(~ industry) +
  labs(title = "Industry wise Yearly % Change & # in employment",
       y = "# Employment (in Millions)") +
  theme(legend.position = "none", panel.grid.major = element_blank(),
        strip.text = element_text(size=8)) +
  guides(x = guide_axis(n.dodge = 2)) +
  
  geom_text(aes(x = year, y = employ_change, label = paste0(round(employ_change*100,
                                                                  digits=1),"%")
                , col = employ_change > 0), 
            nudge_y = -3000000, size = 2.2, angle = 45)

ggsave("actual_perc_change_in_emp.jpg")

adding geom_text to geom_line

employed_ind_cleaned %>% 
  group_by(industry, year) %>% 
  summarise(employment_yrwise = sum(employ_n)) %>% 
  mutate(employ_change = (employment_yrwise - lag(employment_yrwise))/
           lag(employment_yrwise) ) %>% 
  na.omit() %>%
  mutate(industry = str_wrap(industry, width = 20)) %>% 
  
  ggplot() +
  geom_col(aes(x = year, y = employ_change * 20000000, fill = employ_change > 0), alpha = 0.4) +
  geom_line(aes(x = year, y = employment_yrwise), group =1) +
  geom_point(aes(x = year, y = employment_yrwise, col = employ_change > 0)) +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  
  facet_wrap(~ industry) +
  labs(title = "Industry wise Yearly % Change & # in employment",
       y = "# Employment (in Millions)") +
  theme(legend.position = "none", panel.grid.major = element_blank(),
        strip.text = element_text(size=8)) +
  guides(x = guide_axis(n.dodge = 2)) +
  
  geom_text(aes(x = year, y = (employ_change * 20000000) - 3000000, 
                label = paste0(round(employ_change*100,
                                     digits=1),"%")
                , col = employ_change > 0), 
            size = 2, angle = 45) +
  
  geom_text(aes(x = year, y = employment_yrwise + 5000000, 
                label = paste(round(employment_yrwise/1000000, digits=1), "M")
                , col = employ_change > 0), 
            size = 2, angle = 35)

ggsave("actual_perc_change_in_emp2.jpg")

3.8 Occupation Emp Comparison

3.8.1 Major Occupation

employed_ind_cleaned %>% 
  
  ggplot(aes(x = year, y = employ_n, fill = major_occupation)) +
  geom_col() +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  # theme(legend.position = "right", legend.direction = "vertical") +
  labs(title = "Number of employs in Major occupation, Industry wise across the years",
       subtitle = "Keeping scale fixed for industry level comparison") +
  facet_wrap(~ industry)

employed_ind_cleaned %>% 
  
  ggplot(aes(x = year, y = employ_n, fill = major_occupation)) +
  geom_col() +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme(legend.position = "none",
        strip.text = element_text(size=7)) +
  labs(title = "Number of employs in Major occupation, Industry wise across the years",
       subtitle = "Keeping scale fixed for industry level comparison") +
  facet_wrap(~ major_occupation)

3.8.2 Minor occupation

employed_ind_cleaned %>% 
  
  ggplot(aes(x = year, y = employ_n, fill = minor_occupation)) +
  geom_col() +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  # theme(legend.position = "right", legend.direction = "vertical") +
  labs(title = "Number of employs in Minor occupation, Industry wise across the years",
       subtitle = "Keeping scale fixed for industry level comparison") +
  facet_wrap(~ industry)

employed_ind_cleaned %>% 
  mutate(minor_occupation = str_wrap(minor_occupation, width = 25)) %>% 
  
  ggplot(aes(x = year, y = employ_n, fill = minor_occupation)) +
  geom_col() +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme(legend.position = "none",
        strip.text = element_text(size=8)) +
  labs(title = "Number of employs in Minor occupation, Industry wise across the years",
       subtitle = "Keeping scale fixed for industry level comparison") +
  facet_wrap(~ minor_occupation)

3.8.3 Major & Minor

employed_ind_cleaned %>% 
  mutate(major_occupation = str_wrap(major_occupation, width = 30)) %>%
  
  ggplot(aes(x = year, y = employ_n, fill = minor_occupation)) +
  geom_col() +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme(strip.text = element_text(size=8)) +
  labs(title = "Number of employs in Minor occupation, Major occupation wise across the years",
       subtitle = "Keeping scale fixed for industry level comparison") +
  facet_wrap(~ major_occupation)

3.9 Gender Industry Comparison

employed %>% 
  na.omit() %>% 
  filter(dimension == "Gender") %>% 
  
  ggplot(aes(x = year, y = employ_n, fill = race_gender)) +
  geom_col() +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme(strip.text = element_text(size=7)) +
  labs(title = "Number of employs industry wise across the years",
       subtitle = "Colored by Gender") +
  facet_wrap(~ industry) +
  scale_fill_tableau()

employed %>% 
  na.omit() %>% 
  filter(dimension == "Gender") %>% 
  mutate(industry = fct_lump(industry, 11, w = employ_n)) %>% 
  group_by(year, industry, race_gender) %>% 
  summarise(employ_n = sum(employ_n)) %>% 
  
  ggplot(aes(x = year, y = employ_n, col = race_gender)) +
  geom_line(size = 0.9) +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme_bw() +
  theme(panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(),
        strip.text = element_text(size=7),
        legend.position = "top") +
  labs(title = "Number of employs industry wise based on Gender across the years") +
  facet_wrap(~ industry) +
  guides(x = guide_axis(n.dodge = 3)) +
  scale_color_tableau()

employed %>% 
  na.omit() %>% 
  filter(dimension == "Gender") %>% 
  mutate(industry = fct_lump(industry, 11, w = employ_n)) %>% 
  group_by(year, industry, race_gender) %>% 
  summarise(employ_n = sum(employ_n)) %>% 
  
  ggplot(aes(x = year, y = employ_n, col = race_gender)) +
  geom_line(size = 0.9) +
  scale_y_log10(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme_bw() +
  theme(panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(),
        strip.text = element_text(size=7),
        legend.position = "top") +
  labs(title = "Log of employs industry wise based on Gender across the years") +
  facet_wrap(~ industry) +
  guides(x = guide_axis(n.dodge = 3)) +
  scale_color_tableau()

employed %>% 
  na.omit() %>% 
  filter(dimension == "Gender") %>% 
  mutate(industry = fct_lump(industry, 15, w = employ_n)) %>% 
  group_by(year, industry, race_gender) %>% 
  summarise(employ_n = sum(employ_n)) %>% 
  
  ggplot(aes(x = year, y = employ_n, col = race_gender)) +
  geom_line(size = 0.9) +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  theme_bw() +
  theme(panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(),
        strip.text = element_text(size=7),
        legend.position = "top") +
  labs(title = "Number of employs industry wise based on Gender across the years",
       subtitle = "(Free scale comparison)") +
  facet_wrap(~ industry, scales = "free_y") +
  guides(x = guide_axis(n.dodge = 3)) +
  scale_color_tableau(guide = guide_legend(reverse = TRUE))

3.10 2019-2020 Emply change

compare_2019_2020 <- employed %>% 
  na.omit() %>% 
  filter(year %in% c(2019, 2020)) %>% 
  arrange(year) %>% 
  group_by(industry, year, dimension, race_gender) %>% 
  summarise(employ_n = sum(employ_n)) %>% 
  
  group_by(industry, dimension, race_gender) %>% 
  summarise(ratio = last(employ_n) / first(employ_n),
            change = ratio -1,
            employed_2019 = first(employ_n),
            employ_2020 = last(employ_n)) %>%
  mutate(industry = fct_reorder(industry, change, sum)) %>% 
  ungroup()

compare_2019_2020

compare_2019_2020 %>% 
  filter(dimension == "Total") %>% 
  mutate(industry = fct_reorder(industry, change)) %>% 
  
  ggplot(aes(x = change, y = industry, fill = industry)) +
  geom_col() +
  theme(legend.position = "none") +
  scale_x_continuous(labels = percent_format()) +
  labs(title = "Industry %Change in emply. from 2019 to 2020",
       y = "") 

compare_2019_2020 %>% 
  filter(dimension == "Total") %>% 
  mutate(industry = fct_reorder(industry, change)) %>% 
  
  ggplot(aes(x = change, y = industry, fill = industry)) +
  geom_col() +
  theme(legend.position = "none") +
  scale_x_continuous(labels = percent_format()) +
  labs(title = "Industry %Change in emply. from 2019 to 2020",
       y = "") +
  geom_label(aes(label = employ_2020), size = 3, color = "white")

compare_2019_2020 %>% 
  filter(dimension == "Gender") %>% 
  mutate(industry = fct_reorder(industry, change)) %>% 
  
  ggplot(aes(x = change, y = industry, fill = race_gender)) +
  geom_col() +
  theme(legend.position = "right", legend.direction = "vertical") +
  scale_x_continuous(labels = percent_format()) +
  scale_fill_tableau() +
  labs(title = "Industry %Change in emply. from 2019 to 2020",
       y = "")

compare_2019_2020 %>% 
  filter(dimension == "Gender") %>% 
  mutate(industry = fct_reorder(industry, change)) %>% 
  
  ggplot(aes(x = change, y = industry, fill = race_gender)) +
  geom_col(position = "dodge") +
  theme(legend.position = "right", legend.direction = "vertical") +
  scale_x_continuous(labels = percent_format()) +
  scale_fill_tableau(guide = guide_legend(reverse = TRUE)) +
  labs(title = "Industry %Change in emply. from 2019 to 2020",
       y = "")

3.10.1 lollypop plot

3.10.1.1 Gender

compare_2019_2020 %>% 
  filter(dimension == "Gender") %>% 
  mutate(industry = fct_reorder(industry, change)) %>% 
  
  ggplot(aes(x = change, y = industry, col = race_gender)) +
  geom_errorbarh(aes(xmin = 0, xmax = change), height = 0) +
  geom_point(aes(size = employed_2019)) +
  theme(legend.position = "right", legend.direction = "vertical") +
  scale_x_continuous(labels = percent_format()) +
  scale_color_tableau(guide = guide_legend(reverse = TRUE)) +
  scale_size_continuous(guide = FALSE) +
  labs(title = "Industry %Change in emply. from 2019 to 2020",
       y = "", col = "Gender", size = "2019 employ #")

compare_2019_2020 %>% 
  filter(dimension == "Gender") %>% 
  mutate(industry = fct_reorder(industry, change)) %>% 
  
  ggplot(aes(x = change, y = industry, col = race_gender)) +
  
  geom_errorbarh(aes(xmin = 0, xmax = change), height = 0,
             position = position_dodge(width = .6)) +
  geom_point(aes(size = employed_2019),
             position = position_dodge(width = .6)) +
  geom_vline(xintercept = 0, lty = 2, size = 1) +
  
  theme(legend.position = "top",
        panel.grid.major = element_blank()) +
  scale_x_continuous(labels = percent_format()) +
  scale_color_tableau(guide = guide_legend(reverse = TRUE)) +
  scale_size_continuous(guide = FALSE) +
  expand_limits(x = .2) +
  
  labs(title = str_wrap("% Change in Emply. for Industries", 35),
       subtitle = "(from: 2019 to 2020) \n \nSize is proportional to emply # in 2019 \n Lollypop Respresents Gender",
       caption = "Created by ViSa",
       y = "", x = "Change in employment from 2019-2020",
       col = "Gender" )

ggsave(filename = "Industry-gender-lolypop.jpg")

3.10.1.2 Race

compare_2019_2020 %>% 
  filter(dimension == "Race") %>% 
  mutate(industry = fct_reorder(industry, change)) %>% 
  
  ggplot(aes(x = change, y = industry, col = race_gender)) +
  
  geom_errorbarh(aes(xmin = 0, xmax = change), height = 0,
             position = position_dodge(width = .6)) +
  geom_point(aes(size = employed_2019),
             position = position_dodge(width = .6)) +
  geom_vline(xintercept = 0, lty = 2, size = 1) +
  
  theme(legend.position = "top",
        panel.grid.major = element_blank()) +
  scale_x_continuous(labels = percent_format()) +
  scale_color_tableau(guide = guide_legend(reverse = TRUE)) +
  scale_size_continuous(guide = FALSE) +
  
  labs(title = str_wrap("% Change in Emply. for Industries", 35),
       subtitle = "(from: 2019 to 2020) \n \nSize is proportional to emply # in 2019 \n Lollypop Respresents Race",
       caption = "Created by ViSa",
       y = "", x = "Change in employment from 2019-2020", 
       col = "Race")

ggsave(filename = "Industry-race-lolypop.jpg")

3.10.2 major occupation

3.10.2.1 Gender

employed %>% 
  na.omit() %>% 
  filter(year %in% c(2019, 2020)) %>% 
  arrange(year) %>% 
  group_by(year, dimension, race_gender, major_occupation) %>% 
  summarise(employ_n = sum(employ_n)) %>% 
  
  group_by(major_occupation, dimension, race_gender) %>% 
  summarise(ratio = last(employ_n) / first(employ_n),
            change = ratio -1,
            employed_2019 = first(employ_n),
            employ_2020 = last(employ_n)) %>%
  mutate(industry = fct_reorder(major_occupation, change, sum)) %>% 
  ungroup() %>% 
  
  filter(dimension == "Gender") %>%
  
  mutate(major_occupation = fct_reorder(major_occupation, change)) %>% 
  
  ggplot(aes(x = change, y = major_occupation, col = race_gender)) +
  
  geom_errorbarh(aes(xmin = 0, xmax = change), height = 0,
             position = position_dodge(width = .6)) +
  geom_point(aes(size = employed_2019),
             position = position_dodge(width = .6)) +
  geom_vline(xintercept = 0, lty = 2, size = 1) +
  
  theme(legend.position = "top",
        panel.grid.major = element_blank()) +
  scale_x_continuous(labels = percent_format(), limits = c(-.2,.1)) +
  scale_color_tableau(guide = guide_legend(reverse = TRUE)) +
  scale_size_continuous(guide = FALSE) +
  
  labs(title = str_wrap("% Change in Emply. by Major Occupation", 35),
       subtitle = "(from: 2019 to 2020) \n \nSize is proportional to emply # in 2019 \n Lollypop Respresents Gender",
       caption = "Created by ViSa",
       y = "", x = "Change in employment from 2019-2020", 
       col = "Gender")

3.10.2.2 Race

employed %>% 
  na.omit() %>% 
  filter(year %in% c(2019, 2020)) %>% 
  arrange(year) %>% 
  group_by(year, dimension, race_gender, major_occupation) %>% 
  summarise(employ_n = sum(employ_n)) %>% 
  
  group_by(major_occupation, dimension, race_gender) %>% 
  summarise(ratio = last(employ_n) / first(employ_n),
            change = ratio -1,
            employed_2019 = first(employ_n),
            employ_2020 = last(employ_n)) %>%
  mutate(industry = fct_reorder(major_occupation, change, sum)) %>% 
  ungroup() %>% 
  
  filter(dimension == "Race") %>% 
  
  mutate(major_occupation = fct_reorder(major_occupation, change)) %>% 
  
  ggplot(aes(x = change, y = major_occupation, col = race_gender)) +
  
  geom_errorbarh(aes(xmin = 0, xmax = change), height = 0,
             position = position_dodge(width = .6)) +
  geom_point(aes(size = employed_2019),
             position = position_dodge(width = .6)) +
  geom_vline(xintercept = 0, lty = 2, size = 1) +
  
  theme(legend.position = "top",
        panel.grid.major = element_blank()) +
  scale_x_continuous(labels = percent_format(), limits = c(-.2,.1)) +
  scale_color_tableau(guide = guide_legend(reverse = TRUE)) +
  scale_size_continuous(guide = FALSE) +
  
  labs(title = str_wrap("% Change in Emply. by Major Occupation", 35),
       subtitle = "(from: 2019 to 2020) \n \nSize is proportional to emply # in 2019 \n Lollypop Respresents Race",
       caption = "Created by ViSa",
       y = "", x = "Change in employment from 2019-2020", 
       col = "Race")

3.10.3 minor occupation

3.10.3.1 Gender

employed %>% 
  na.omit() %>% 
  filter(year %in% c(2019, 2020)) %>% 
  arrange(year) %>% 
  group_by(year, dimension, race_gender, minor_occupation) %>% 
  summarise(employ_n = sum(employ_n)) %>% 
  
  group_by(minor_occupation, dimension, race_gender) %>% 
  summarise(ratio = last(employ_n) / first(employ_n),
            change = ratio -1,
            employed_2019 = first(employ_n),
            employ_2020 = last(employ_n)) %>%
  mutate(industry = fct_reorder(minor_occupation, change, sum)) %>% 
  ungroup() %>% 
  
  filter(dimension == "Gender") %>%
  
  mutate(minor_occupation = fct_reorder(minor_occupation, change)) %>% 
  
  ggplot(aes(x = change, y = minor_occupation, col = race_gender)) +
  
  geom_errorbarh(aes(xmin = 0, xmax = change), height = 0,
             position = position_dodge(width = .6)) +
  geom_point(aes(size = employed_2019),
             position = position_dodge(width = .6)) +
  geom_vline(xintercept = 0, lty = 2, size = 1) +
  
  theme(legend.position = "top",
        panel.grid.major = element_blank()) +
  scale_x_continuous(labels = percent_format(), limits = c(-.2,.1)) +
  scale_color_tableau(guide = guide_legend(reverse = TRUE)) +
  scale_size_continuous(guide = FALSE) +
  
  labs(title = str_wrap("% Change in Emply. by Minor Occupation", 35),
       subtitle = "(from: 2019 to 2020) \n \nSize is proportional to emply # in 2019 \n Lollypop Respresents Gender",
       caption = "Created by ViSa",
       y = "", x = "Change in employment from 2019-2020", 
       col = "Gender")

3.10.3.2 Race

employed %>% 
  na.omit() %>% 
  filter(year %in% c(2019, 2020)) %>% 
  arrange(year) %>% 
  group_by(year, dimension, race_gender, minor_occupation) %>% 
  summarise(employ_n = sum(employ_n)) %>% 
  
  group_by(minor_occupation, dimension, race_gender) %>% 
  summarise(ratio = last(employ_n) / first(employ_n),
            change = ratio -1,
            employed_2019 = first(employ_n),
            employ_2020 = last(employ_n)) %>%
  mutate(industry = fct_reorder(minor_occupation, change, sum)) %>% 
  ungroup() %>% 
  
  filter(dimension == "Race") %>% 
  
  mutate(minor_occupation = fct_reorder(minor_occupation, change)) %>% 
  
  ggplot(aes(x = change, y = minor_occupation, col = race_gender)) +
  
  geom_errorbarh(aes(xmin = 0, xmax = change), height = 0,
             position = position_dodge(width = .6)) +
  geom_point(aes(size = employed_2019),
             position = position_dodge(width = .6)) +
  geom_vline(xintercept = 0, lty = 2, size = 1) +
  
  theme(legend.position = "top",
        panel.grid.major = element_blank()) +
  scale_x_continuous(labels = percent_format(), limits = c(-.2,.1)) +
  scale_color_tableau(guide = guide_legend(reverse = TRUE)) +
  scale_size_continuous(guide = FALSE) +
  
  labs(title = str_wrap("% Change in Emply. by Minor Occupation", 35),
       subtitle = "(from: 2019 to 2020) \n \nSize is proportional to emply # in 2019 \n Lollypop Respresents Race",
       caption = "Created by ViSa",
       y = "", x = "Change in employment from 2019-2020", 
       col = "Race")

3.10.4 Industry wise % change

3.10.4.1 Total

library(ggrepel)

compare_2019_2020 %>% 
  filter(dimension == "Total") %>% 
  mutate(industry = fct_reorder(industry, change)) %>% 
  
  ggplot(aes(employed_2019, change)) +
  geom_point() +
  geom_text_repel(aes(label = industry), size = 3) +
  geom_hline(yintercept = 0, lty = 2, col = "red") +
  
  scale_x_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  scale_y_continuous(labels = percent_format()) +
  
  labs(title = "Overall % Emply Change for Industries",
       subtitle = "(in: 2019 to 2020)")

3.10.4.2 Race

compare_2019_2020 %>% 
  filter(dimension == "Race",
         race_gender == "Asian") %>% 
  mutate(industry = fct_reorder(industry, change)) %>% 
  
  ggplot(aes(employed_2019, change)) +
  geom_point() +
  geom_text_repel(aes(label = industry), size = 3) +
  geom_hline(yintercept = 0, lty = 2, col = "red") +
  
  scale_x_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
  scale_y_continuous(labels = percent_format()) +
  
  labs(title = "Emply % Change for Asians in Industry",
       subtitle = "(in: 2019 to 2020)")

3.11 More Exploration

employed %>% 
  na.omit() %>% pull(race_gender) %>% 
  table()

.
                    Asian Black or African American                       Men 
                     1254                      1254                      1254 
                    TOTAL                     White                     Women 
                     1254                      1254                      1254 

  
employed %>% 
  na.omit() %>% pull(dimension) %>% 
  table()

.
Gender   Race  Total 
  2508   3762   1254 

Seems like we dont have data for Gender among Races, so skiping the analysis of combination of both.

employed

employed %>% 
  pull(race_gender) %>% 
  table()

.
                    Asian Black or African American                       Men 
                     1254                      1386                      1386 
                    TOTAL                     White                     Women 
                     1386                      1386                      1386 

3.12 End of this EDA